Is positive EV guaranteed to profit?

No. Positive EV is the long-run expectation, not a per-bet guarantee. A +EV bet at 55% true probability still loses 45% of the time. Profit shows up over hundreds or thousands of bets, not over a weekend. Variance is the price of the edge.

How do you calculate true probability?

Three ways. Build a model that simulates the outcome from inputs (player form, matchup, pace). Use a market consensus de-vigged across many books. Or use a base-rate prior (historical hit rate for the prop type) adjusted for current context. Elite Picks uses the first approach with the third as a sanity check.

What's the difference between EV and ROI?

EV is the expected profit per bet at the moment you place it, calculated from your probability estimate. ROI is the realized profit per dollar wagered after the bets settle. EV is forward-looking and theoretical. ROI is backward-looking and actual. Over a large sample, ROI converges to EV if your probabilities are calibrated.

Does EV account for the vig?

Yes, if you do it right. The implied probability from sportsbook odds includes the book's margin, so you have to de-vig (strip the margin out) before comparing to your model. Skipping that step makes -110 look like a coin flip when it's actually 52.4% implied. The vig is baked into every line. EV math has to back it out.

How does Elite Picks use EV in the EP score?

EV is the raw input. The EP score multiplies edge by model calibration tier and confidence on the specific prediction, then normalizes to a 0-100 scale. A 6% edge from a well-calibrated model with tight prediction bounds scores higher than a 10% edge from a model with loose bounds and recent calibration drift. EV decides which bets we surface. The score decides how we rank them.

What is expected value in sports betting?

Last updated: 2026-05-26

Expected value in sports betting is the average profit or loss you’d realize per bet if you placed the same wager an infinite number of times. It’s calculated as (probability of winning × payout) - (probability of losing × stake). A positive EV bet has a long-run mathematical edge over the sportsbook. A negative EV bet doesn’t, regardless of whether it cashes on any given night. EV is the single number that separates disciplined betting from gambling, and it’s the foundation of every model-driven approach to the markets.

What’s the formula for expected value?

The textbook formula is:

EV = (P_win × profit_if_win) - (P_loss × stake)

Where P_win is your estimated probability the bet hits, P_loss is 1 - P_win, profit_if_win is the payout net of stake, and stake is what you risk. Positive number, you have an edge. Negative, the book does.

A cleaner way to express the same thing in percentage terms:

EV% = (P_win × decimal_odds) - 1

If that’s greater than zero, the bet is +EV.

Both formulas assume one thing. Your probability estimate is honest. If your model says 55% and the true rate is 50%, the math is right but the inputs are wrong, and your “edge” will evaporate over a real sample.

A worked example at -110 odds

Take a standard player prop posted at -110 on both sides. You believe the over hits 55% of the time. Should you bet it?

First, convert -110 to a payout. American odds of -110 mean you risk $110 to win $100, so a $100 stake pays $90.91 in profit. Decimal odds are 1.909.

Now run the EV math:

P_win = 0.55
P_loss = 0.45
Profit if win = $90.91 per $100 risked
Loss if lose = $100

EV = (0.55 × $90.91) - (0.45 × $100) = $50.00 - $45.00 = $5.00

55%Your estimate

52.4%What -110 implies

$5.00Expected profit per $100

-$4.55Per $100 at a true coin flip

That’s $5 of expected profit per $100 risked, or 5% ROI per bet at this edge. Compound it across hundreds of bets at this edge and the math becomes serious money. A bettor placing 500 wagers a season at this edge with $100 units expects $2,500 in long-run profit, before variance smooths in.

Now look at what happens if your probability estimate is wrong. At 52%, EV drops to $0.27 per $100. At 50% (a true coin flip on a -110 line), EV is -$4.55. The book wins. The 4.55% loss rate is exactly the vig at -110/-110, which works out to a 4.76% hold across the two-way market.

Why does EV matter more than win rate?

Win rate is the metric beginners track. It’s also the metric that misleads them.

Both look identical on a Twitter screenshot. Only one prints.

EV collapses win rate and price into a single number. If a bettor tells us they hit 56% of their plays, we can’t say whether they’re sharp. If they tell us their average bet was +4% EV at the time of placement, we know they’re sharp, and we know roughly how much they should be up after a thousand bets.

The clearest demonstration sits in closing line value research from Pinnacle. Bettors whose bets consistently beat the closing line by 2 to 3% post long-run ROI in the same range, regardless of weekly hit rate variance (Pinnacle, “Why CLV matters”). Hit rate over a single month tells you almost nothing. EV at placement tells you almost everything.

EV vs ROI vs CLV: which metric matters when?

Three metrics get conflated in betting writing. They measure different things at different points in the bet lifecycle.

Metric	What it measures	When you see it	Reliability
Expected Value (EV)	Forecasted profit per bet given your probability estimate	At bet placement	Only as good as your model
Closing Line Value (CLV)	How much your number beat the closing price	After the book closes the market	High; correlates with long-run ROI
ROI	Realized profit per dollar wagered	After bets settle	Lagging; noisy over <500 bets

Most bettors invert this order and judge themselves on a week of ROI. That’s how they go broke chasing.

Common mistakes when calculating EV

Three errors show up over and over.

Confusing EV with profit. A +5% EV bet still loses 45% of the time at -110. The EV is what the bet is worth in expectation, not what it pays tonight. We’ve watched sharp bettors abandon +EV strategies after two bad weeks because the dollar P&L didn’t match the projected EV. The math caught up by month four. They quit at week three.

Ignoring the vig. A line of -110 on both sides isn’t a 50/50 priced fairly. It’s a 52.4% implied probability on each side, which sums to 104.8%. That 4.8% gap is the book’s hold. Comparing your model probability to the raw implied probability (without de-vigging) consistently underestimates your edge on overs and overestimates it on unders, or vice versa. Always de-vig first.

Trusting an uncalibrated model. A model that says 60% but actually hits 54% generates wildly inflated EV at every line. Calibration matters more than discrimination for EV math. The MLB home run model we shipped first looked sharp on backtests until we noticed eight features were evaluating to NaN at inference time. Probabilities looked fine. Calibration was wrecked. We caught it on the ledger, not in the model file.

How does Elite Picks use EV?

We build one calibrated model per prop family (NBA points, MLB strikeouts, NFL receiving yards). Each model outputs the full probability distribution per player-game and reads P(over line) directly off the CDF. We de-vig the sportsbook two-way market to get a fair implied probability per side, compare, and compute edge. EV per unit is edge multiplied by payout.

The EV is the raw input to the EP score, which sits at the application layer rather than inside the model itself. The score multiplies edge by the model’s historical calibration tier and by the confidence on this specific prediction (how tight the prediction interval is). Plays at EP 90 and above carry both the highest edge and the cleanest closing-line value on our internal ledger. Plays below EP 75 don’t surface at all, even if the raw EV looks attractive.

That last filter matters. Raw EV ranks bets by mathematical edge. The EP score ranks them by edge that we trust enough to recommend. The difference is calibration, and it’s the difference between a model and a research tool.

Where to go next

The companion concepts to EV are implied probability (how sportsbook odds convert to probabilities), vig (the book’s built-in margin), and closing line value (the leading indicator that proves your probabilities were calibrated). The Elite Picks methodology page walks through how we go from raw data to a calibrated probability to an EV number to a ranked recommendation.

Elite Picks is a research and analytics platform. We do not accept wagers, hold gambling licenses, or facilitate any real-money transactions. All content is for informational and research purposes.